Download From: AAAI Technical Report S-9 -0 . Compilation copyright © 199

From: AAAI Technical Report WS-98-07. Compilation copyright © 1998, AAAI (www.aaai.org). All rights reserved.
PREFACE
Manydream of being able to predict the future. In finance, accurate predictions can direct
portfolio management decisions. In marketing, knowledge of future demand for products and
services can direct capital allocation. In operations support, predicting future problems (such as
equipment failures and network congestion) can reduce or avoid their associated costs. In order to
predict the future, we try to understand the past by analyzing historical data to discover patterns
that can inform us about the processes being studied.
Each year, new opportunities open up for temporal data mining as more time-series data become
available from high-profile applications. Investment firms are spending large sums of moneyon
research attempting to predict the behaviors of individual financial instruments, as well as entire
markets. The telecommunications industry is relying increasingly on performance monitoring,
hoping to improve network reliability
by catching failures before they occur. Workon fraud and
intrusion detection often involves describing behavior trends. Researchers in data mining on the web
are realizing the importance of dealing with changing information content and drifting preferences
of their users.
This recent increase of interest in time series problems led us to organize this workshop, Predicting the Future: AI Approaches to Time-series Problems, held on July 27, 1998 in conjunction
with the Fifteenth National Conference on Artificial Intelligence (AAAI-98). The purpose of the
workshop was to gather together AI researchers studying various aspects of time-series analysis.
Our hope was to reach some commonground in an area of research that is receiving increasing
attention, as well as to discuss new results. These working notes contain the technical papers
presented at the workshop.
We thank the authors, attendees, and invited speakers for their efforts and enthusiasm in
making this possible. Weare indebted to AAAIfor organizational and funding assistance, and for
publishing these working notes; to David Leake, Chair of the AAAI-98WorkshopCommittee; and
to our anonymousworkshop-proposal reviewers for their suggestions and encouragement.
Andrea Danyluk, Tom Fawcett, and Foster Provost
iii